多格式音频感知哈希算法-北京邮电大学学报.PDF

2016年8月 北 京 邮 电 大 学 学 报 Aug.2016 第39卷 第4期 Journal of Beijing University of Posts and Telecommunications Vol.39 No.4 摇 摇 文章编号:1007鄄5321(2016)04鄄0077鄄06 DOI:10.13190/ j.jbupt.2016.04.015 多格式音频感知哈希算法 张秋余,摇 省鹏飞,摇 黄羿博,摇 董瑞洪,摇 杨仲平 (兰州理工大学 计算机与通信学院,兰州730050) 摘要:提出一种基于双树复小波变换的多格式音频感知哈希算法,解决了现有音频认证算法音频格式单一、算法不 通用、效率低的问题. 首先对预处理后的音频信号进行全局双树复小波变换,获得信号的实小波和复小波系数,对 它们分别分帧,帧数相同;对实小波系数计算每帧信号Teager能量算子的模值,作为实小波系数的帧间特征,接着 对每帧信号再分帧,提取再分帧帧信号的短时能量作为实小波系数的帧内特征;对复小波系数求取每帧信号的熵 值作为复小波系数的帧间特征;最后对上述特征分别进行哈希构造,生成感知哈希序列. 实验结果表明,该算法对 5种不同格式的音频都具有强鲁棒性,且区分性好,效率高,并能实现小范围篡改检测. 关摇 键摇 词:音频认证;多格式音频;感知哈希;双树复小波变换;篡改检测 中图分类号:TN912郾3;TP309郾7摇 摇 摇 摇 文献标志码:A Perceptual Hashing Algorithm for Multi鄄Format Audio ZHANG Qiu鄄yu,摇 XING Peng鄄fei,摇 HUANG Yi鄄bo,摇 DONG Rui鄄hong,摇 YANGZhong鄄ping (School of Computer and Communication,Lanzhou University of Technology,Lanzhou730050,China) Abstract:A novel multi鄄format audio perceptual hashing algorithm based on dual tree complex wavelet transform (DT鄄CWT) was proposed. It solves the problems of the existing audio authentication algo鄄 rithms,including that audio filesarekept in a singleformat,andalgorithmsarenot genericandlow effi鄄 ciency. The proposed algorithm first applies the global DT鄄CWT to the audio signal after pre鄄processing conducts to obtain the real and complex wavelet coefficients. Next,the coefficients are partitioned in some frames respectively,andtheframenumbersare same. Fortherealwavelet coefficients,themodule values of teager energy operator in everyframearecomputedto serveasitsinter鄄framefeature. Andthen short鄄time energy ofthenew signal,whichisgeneratedtoframetheframesignal,iscomputedtoserveas itsintra鄄framefeature. Forthecomplexwavel